Meta released Metaspirit LM, an open-source multimodal model that handles text and speech like OpenAI's Advanced Voice. It's free to study but not for commercial use yet, and the audio quality needs work, but the open release could help the community build better competitors.
Oh, that Meta AI just released their own GPT4 Omni advanced voice competitor. Check it out, they're calling this Metaspirit LM. It's their first open source multimodal language model that freely mixes text and speech just like Advanced Voice mode. But it being open source, the community at large can actually learn a lot more from it than they would learn from Advanced voice mode because OpenAI doesn't tell us anything about about how they made it. Unfortunately for now it's still non commercial open source which means you can't make money off. And here's what it sounds like right now. There was a beautiful cat named Peter. His mother taught him how to climb a tree. I feel very excited about the upcoming trip to Paris. So yeah, it is no OpenAI advanced voice, that's for sure. But hey, at least the community's gonna put this one to good use and it's going to help develop future models that are actually competitive.
